fabfile.py has gotten crufty

[divergentdave]

The current state of the fabfile doesn't totally make sense right now, because the scrapers and web server are now on different servers, and other changes. We haven't touched the frontend in quite a while, so it hasn't come up yet.

• Currently the fabfile will log in to every web server. Either there's just one web server, so that makes sense, or there's a production server and a staging server, in which case that isn't very useful.
• The reindex function won't work anymore, because it would be run on the scraper, which doesn't have the data EBS volume mounted. If we can have different fabric commands target different servers, this would be a good place to do so.
• It might make sense to drop the "cold" variants of commands, since the user-data scripts currently start up servers with forever, so there's no cold start to be done.
• By rights, there ought to be a call to rake elasticsearch:init_dashboard, either under reindex or somewhere else.

[konklone]

Should we just delete it entirely? How are you running deploys at the moment?

[divergentdave]

All deploys in recent memory have been scrapers only or infrastructure related, not frontend, so I've been using the webhook to update scrapers and spinning up fresh servers when that's needed. I think it's worth keeping around for the next time there's frontend work to deploy.